Body adiposity index and all-cause and cardiovascular disease mortality in men

Objective To evaluate the association of body adiposity index (BAI) with all-cause and cardiovascular disease (CVD) mortality risk. Design and Methods The current analysis comprised 19 756 adult men who enrolled in the Aerobics Centre Longitudinal Study and completed a baseline examination during 1988-2002. All-cause and CVD mortality was registered till December 31, 2003. Results During an average follow-up of 8.3 years (163 844 man-years), 353 deaths occurred (101 CVD deaths). Age- and examination year-adjusted hazard ratios (HRs) and 95% confidence intervals (95% CIs) for all-cause mortality risk were higher for men with high values of BMI (HR = 1.63, 95% CI = 1.19–2.23), waist circumference (1.55, 1.22-1.96) and percentage of body fat (%BF) (1.36, 1.04-1.31), but not for men with high values of BAI (1.28, 0.98-1.66). The HRs for CVD mortality risks were higher for men with high values in all adiposity measures (HRs ranged from 1.73 to 2.06). Most of these associations, however, became nonsignificant after adjusting for multiple confounders including cardiorespiratory fitness. Conclusion BAI is not a better predictor of all-cause and CVD mortality risk than BMI, waist circumference or %BF.


Introduction
The prevalence of obesity among adults has reached epidemic levels worldwide (1,2). It is well known that overweight and obesity are major risk factors that can lead to further morbidity and mortality (3)(4)(5).
BMI and waist circumference are strong predictors of obesity-related morbidity and mortality (6)(7)(8). Despite their limitations, both are commonly used as adiposity measures in large epidemiological studies where the use of more accurate methods such as dual-energy X-ray absorptiometry (DXA), hydrostatic weighing, bioelectrical impedance, or even skinfold thickness are limited because of its complexity and/or cost (9,10).
To our knowledge, the association between BAI and mortality has not been investigated. Such a study is needed to determine the potential use of BAI as a mortality risk predictor in large epidemiological studies. Therefore, the purpose of this study was to evaluate the association of BAI with all-cause and CVD mortality risk. In addition, we analyzed the cross-sectional associations between BAI and traditional CVD risk factors in a sample of men participating in the Aerobics Center Longitudinal Study (ACLS).

Methods and Procedures
Subjects Data for this report were from the ACLS, a prospective epidemiological study of individuals who received extensive preventive medical examinations at the Cooper Clinic in Dallas, Texas. Details of the study design and the characteristics of the cohort have been reported previously (24). Study participants were referred by their employers or physicians or were self-referred. They were mainly Caucasian, relatively well-educated, and from middle-to-upper socioeconomic strata. After receiving complete information about the aims and methods of the study, all participants gave written informed consent for the examinations and follow-up. The study protocol was reviewed and approved annually by the Cooper Institute's Institutional Review Board.
For the present analysis, we included all men who received at least one medical examination between 1988 and 2002, and with valid data for hip circumference. For men attending more than one examination we used the first examination (baseline). Among 23,126 men aged !20 years at baseline, we excluded 299 with a history of myocardial infarction or stroke; 1,265 reporting cancer; 28 with BMI <18.5 kg/m 2 ; and 631 not reaching 85% of their age-predicted maximal hart rate (220 minus age in years) on a maximal treadmill exercise test. In addition, 1,147 men with less than one year of mortality follow-up were excluded to minimize potential bias because of serious underlying illness on mortality. The final sample comprised 19,756 men for analyses of all-cause and CVD mortality.

Clinical examination
The clinical examination procedures are described in detail elsewhere (24,25). Briefly, weight, height, and waist and hip circumferences were measured with standard clinical scale, stadiometer, and plastic tape according to ACLS standard procedures. BMI was computed as weight in kilograms divided by height in meters squared (kg/m 2 ), and BAI was calculated according to Bergman et al.'s equation ((hip circumference in cm/height in meters 1.5 ) À 18) (11). %BF was assessed using hydrostatic weighing, the sum of seven skinfold measures, or both methods, following standardized protocols (26). Detailed description of our hydrodensitometry procedures have been published elsewhere (27). Men were classified according to adiposity categories using standard clinical definitions for BMI (18.5-24.9 as normal weight; 25-29.9 as overweight; !30 as obese) and waist circumference (!102 cm as central obese) (28). Because there is not a specific agreement about obesity cut-off points for BAI and %BF, specific tertiles from this population were used.
Resting blood pressure was measured in the seated position by trained technicians using auscultatory methods with mercury sphygmomanometer and was recorded as the first and fifth Korotkoff sounds after !5 minutes of sitting quietly. Two readings separated by 2 minutes were averaged. If the first 2 readings differed by >5 mm Hg, additional readings were obtained and averaged. Hypertension was defined as systolic blood pressure !140 mm Hg, diastolic blood pressure !90 mm Hg, or a history of physician diagnosis. Electrocardiogram (ECG) was measured at rest and with exercise, and abnormal ECG responses included rhythm and conduction disturbances and ischemic ST-T wave abnormalities. Serum samples were analyzed for glucose and total cholesterol using standardized automated bioassays at the Cooper Clinic Laboratory. Diabetes mellitus was defined as fasting plasma glucose concentration of !126 mg/dL, a history of physician diagnosis, or insulin use. Hypercholesterolemia was defined as total cholesterol of !240 mg/dL or a history of physician diagnosis.
Physical activity habits, cigarette smoking habit, alcohol intake, and parental history of CVD were assessed by self-report on a medical history questionnaire. Physically inactive was defined as reporting no leisure-time physical activity in the 3 months before the examination. Smoking status was classified as current smoker or noncurrent smoker. Heavy alcohol consumption was established as >14 units/week. One unit of alcohol intake was defined as a bottle or can of beer [355 mL (12 oz)], a glass of wine [148 mL (5 oz)], or 44 mL (1.5 oz) of hard liquor. Parental history of CVD was defined as the occurrence of heart attacks, coronary disease, angioplasty, or stroke before the age of 50 years in either father or mother.
Cardiorespiratory fitness (CRF) was quantified as the total time of a symptom-limited maximal treadmill exercise test, using a modified Balke protocol (29,30). Total treadmill endurance time (in minutes) on this protocol correlates highly with measured maximal oxygen uptake in men (r ¼ 0.92) (31). To standardize exercise performance, we estimated maximal metabolic equivalents (METs; 1 MET ¼ 3.5 mL O 2 uptake/ kg/min) from the final treadmill speed and grade (32).

Vital status
All participants were followed for mortality from the baseline examination to the date of death or to December 31, 2003. Deaths were identified from the National Center for Health Statistics National Death Index and official death certificates from the departments of vital records of the various states. The underlying cause of death was determined by a nosologist according to the International Classification of Diseases, Ninth Edition, with CVD defined as codes 390-449.9 before 1999 and Tenth Edition, with CVD defined as codes I00-I78 during 1999-2003. The National Death Index has been shown to be an accurate method of ascertaining deaths in observational studies, with high sensitivity (96%) and specificity (100%) (33).

Statistical analysis
Descriptive analyses summarized baseline characteristics of the participants based on vital status and BAI tertiles. Differences between groups were tested using analysis of the variance (ANOVA) for continuous variables and chi-square tests for categorical variables. Partial correlations between body adiposity measures were calculated after controlling for age and examination year. We also examined the association of body adiposity measures with CVD risk factors using linear regression, controlling for age and examination year. Cox proportional hazards regression analysis (timescale: years) was used to estimate mortality rates (deaths per 10,000 man-years of follow-up), hazard ratios (HRs), and associated 95% confidence intervals (95% CIs) for all-cause and CVD mortality, according to adiposity exposure categories from BAI, BMI, waist circumference, and %BF. The lowest adiposity category was used as the reference category. To allow comparisons between adiposity measures, we also presented the results as standardized HRs by transforming each variable to have a mean of 0 and a SD of 1. In multivariable analyses, model 1 accounted for age and examination year. Model 2 included physical activity, smoking habit, alcohol intake, abnormal ECG, hypercholesterolemia, diabetes, hypertension, and parental history of CVD as additional confounders. Model 3 additionally adjusted for CRF. The proportional hazards assumption was examined by comparing the cumulative hazard plots grouped on exposure; no appreciable violations were noted. All the analyses were performed using PASW statistical package version 18.0 (SPSS Inc, Chicago, IL, USA), considering P < 0.05 as being statistically significant.

Results
During an average follow-up of 8.3 years and 163,844 man-years of observation, 353 deaths occurred (101 because of CVD). The characteristics of the study population by vital status and BAI tertiles are shown in Table 1. Decedents presented significantly higher values at baseline for age, %BF, total cholesterol, fasting blood glucose, systolic blood pressure, as well as higher prevalence of abnormal ECG and hypertension. On the contrary, decedents had significantly lower values for CRF. All values and prevalence rates showed significant differences among BAI tertiles, except for the prevalence of heavy drinkers.
Partial correlations examining the association among body adiposity measurements after controlling for age and examination year are shown in Table 2. Although all body adiposity measurements were positively correlated (P 0.001), %BF was more strongly correlated with waist circumference (r ¼ 0.77) and BMI (r ¼ 0.72) than BAI (r ¼ 0.65). Table 3, all body adiposity measures were significantly associated with CVD risk factors (all P < 0.001). Except for systolic blood pressure, correlations between BAI and CVD risk factors were slightly weaker than for BMI, waist circumference, and %BF. Among all the risk factors, CRF showed the strongest association with all adiposity measurements.      Table 5 shows death rates and HRs for CVD mortality. In model 1, CVD mortality risk was higher for men in the highest category of all adiposity measurements (HRs ranged from 1.73 to 2.06). After additional adjustments (model 2 and model 3), the association between adiposity status and risk of CVD mortality became not significant.

According to
To further evaluate the possible bias because of subclinical disease at baseline, all the analyses were repeated excluding participants with resting or exercise abnormal ECG, and also excluding deaths that occurred during the first 3 years of follow-up. The results did not substantially change (data not shown).

Discussion
The results of the present report suggest that, among men enrolled in the ACLS between 1988 and 2002, BAI is not a better predictor for all-cause and CVD mortality risk than other body adiposity measures such as BMI, waist circumference, or %BF. Furthermore, associations between BAI and traditional CVD risk factors (e.g., cholesterol, glucose, blood pressure, and CRF) are slightly weaker than for BMI, waist circumference, or %BF.
The BAI has recently been proposed by Bergman et al. (11) to provide valid direct estimates of %BF. The mean intra-individual difference between BAI and %BF in our study was 3.5%, 2.7%, and 1.9% in lower, middle, and upper BAI tertiles, respectively. Moreover, %BF showed a relatively higher correlation with BMI (r ¼ 0.72) and waist circumference (r ¼ 0.77) than BAI (0.65). Similar results have been found in some BAI validation studies when correlation analyses were sex-stratified, indicating that calculation of BAI would result in less accurate estimates of %BF (12,(17)(18)(19).
Previous studies have showed a strong association of CVD risk factors with traditional body adiposity measures (4,6,34,35), but the association with BAI seemed to be weaker (13,17,18,(20)(21)(22)(23). In fact, only one study performed in 13 women aged 33.6 6 11.5 years with familial partial lipodystrophy found BAI to be more strongly correlated with leptin than BMI (r ¼ 0.57 and r ¼ 0.02, respectively) (15). Our results are consistent with previous findings reporting significant associations between all adiposity measures and CVD risk factors (e.g., cholesterol, glucose, blood pressure, and CRF), and showing slightly weaker associations with BAI. Therefore, BAI seems not to provide a meaningful alternative to traditional adiposity measurements as a CVD risk indicator. The association of obesity with all-cause and CVD mortality is well established (7,8,36,37). To the best of our knowledge, no previous studies have analyzed the ability of BAI to predict all-cause and CVD mortality. According to our results, BAI is not a good predictor of mortality risk. Indeed, only men in the upper BAI tertile showed statistically higher CVD mortality risk after adjusting for age and examination year. However, men in the highest adiposity categories for BMI, waist circumference, and %BF presented statistically higher risks of all-cause and CVD mortality after adjusting for age and examination year. The lack of accuracy of BAI estimating %BF, and the fact that it does not reflect fat distribution, which is known to have a large influence on mortality risk in men (android obesity), could partially explain these differences between mortality and the analyzed body adiposity measurements.
Several studies provide evidence that CRF substantially modifies the association of adiposity measures with mortality (38), but no previous study using BAI included CRF in their analysis. We analyzed three different sets of confounders, and in the fully adjusted model including CRF all the associations between obesity and mortality became not significant.
The results of the present analysis should be interpreted with caution because of several limitations. First, because our study only included Caucasian, well-educated men from middle-to-upper socioeconomic strata, the results may not be extended to other populations. However, the homogeneity of the sample enhances internal validity of our findings because it reduces the likelihood of confounding by these characteristics. Second, %BF was estimated from skinfold measures or hydrostatic weighing, each of which has methodological limitations (39). Third, as we only compared baseline data for exposures, we do not know whether changes in any of these variables occurred during follow-up and how that might have influenced the results. However, we recently published a report on changes in weight, %BF, and CRF with all-cause and CVD mortality (40). When each of these exposures were adjusted for all other exposures, only changes in CRF were associated with mortality. Fourth, the small number of deaths during the follow-up period restricted the statistical power of the analyses. Finally, we could not take into account dietary factors because of lack of adequate dietary information. Despite these limitations, the main strengths of this study included the large, well-characterized cohort of men, the prospective design of the study, the use of two different mortality outcomes, the extensive follow-up period, and the extensive baseline examination that reduced the possible bias of subclinical disease.
In conclusion, the findings of the present analysis add information about the newly proposed BAI, suggesting that BAI is not a better predictor of all-cause and CVD mortality risk in men than traditional adiposity measures. Also, our results show that the association of BAI with CVD risk factors is relatively weaker than for BMI, waist circumference, and %BF. Further epidemiological studies examining the utility of BAI in other populations and women are still needed for a better understanding of the validity of this new index.O